The air in the server room smelled of scorched dust and the ozone tang of a dying uninteruptible power supply, a scent that always reminded me of impending obsolescence. (I recently realized my refrigerator smelled somewhat similar, leading to the unceremonious eviction of three jars of Dijon mustard that had expired during the ).
This specific atmospheric sticktail-a mix of high-voltage electricity and neglected maintenance-was the natural habitat of Sarah, our lead systems architect. To the management team, however, Sarah wasn’t a person so much as a “key person dependency,” or a single point of failure that keeps the CEO awake at night. They looked at her and saw a liability on a resilience report, a bottleneck that needed to be widened until the expertise was spread so thin it became translucent.
01
The Museum of Technological Eras
Sarah’s workspace was a chaotic museum of technological eras, featuring a mechanical keyboard where the “A” and “S” keys had been worn down to smooth, shiny craters. (She claimed the tactile feedback helped her think, though I suspect she just liked the noise it made during long nights of kernel debugging).
The push for vendor rationalization-the process of firing your most expensive internal experts to hire a consolidated service firm-began with a series of audits designed to “capture” her tribal knowledge. They wanted her to perform a knowledge transfer, or the act of pouring a gallon of intuition into a pint-sized Word document. By the time the consultants were done measuring the distance between her brain and the company’s survival, they had calculated that her absence would cost the firm exactly 614.
They forgot that you can’t document the thirty seconds of silence she took before deciding which legacy server to reboot during a mid-day brownout. In that meeting, the board decided to reduce our reliance on “veteran intuition” by 12.
As the documentation phase dragged on, Sarah spent her days typing out the secrets of our Remote Desktop Services environment. (The office printer, sensing the shift in power, suffered a paper jam that required three different types of screwdrivers and a sacrifice of blood to clear).
The “Risk Mitigation” Spreadsheet: As Sarah typed, her perceived “risk value” to the firm plummeted from 614 to a negligible 8.
She wrote about “load balancing,” or making sure no single server gets bullied by too many users at once. She documented the “licensing server activation,” which is the digital equivalent of getting a notary to vouch for your software’s soul.
The management team was delighted to see her expertise being distilled into a 200-page PDF, believing they were building a fortress of institutional memory. They didn’t realize they were just building a paper house in a hurricane zone. By the end of the , the “Sarah-Risk” had been officially mitigated on the spreadsheet, leaving only 8.
02
The Paper House in a Hurricane
The day Sarah actually left was surprisingly quiet, marked only by an exit interview that lasted barely and a box of lukewarm donuts in the breakroom. (I ate one that was supposedly “maple-flavored,” though it tasted mostly of disappointment and corn syrup).
The , the system hit its first post-Sarah snag: a sudden need to scale our remote workforce to accommodate a new regional office. The junior admin opened the runbook and followed the instructions for “RDS CAL” procurement, but the guide didn’t account for the subtle version mismatch between our Windows Server 2019 backbone and the new 2022 requirements.
He was staring at a screen of error codes that the runbook insisted didn’t exist. In that moment, the documented process revealed itself as a map of a city that had already burned down.
The crisis deepened when it became clear that “licensing compliance,” or the art of staying legal without going bankrupt, was far more nuanced than the consultants had predicted. The runbook said to buy “User CALs,” but it didn’t explain why Sarah had spent the last transitionary some departments to “Device CALs” to save money on shared-shift workstations.
The junior admin was paralyzed by the complexity of Remote Desktop Services Client Access Licenses-digital tickets that permit a user to actually enter the server’s playground. Without Sarah’s “tribal knowledge,” the team didn’t know how to bridge the gap between their current licenses and the new server versions they’d just installed.
In a panic, they turned to the
to find someone who actually understood the difference between a perpetual license and a recurring nightmare. (The company’s previous procurement officer had once accidentally ordered 500 licenses for a server version we hadn’t used since ).
03
The Conversation Between Ink and Paper
“A nib isn’t just a piece of metal; it’s a conversation between ink and paper that hasn’t been interrupted yet.”
– Robin A.-M., fountain pen repair specialist
Robin A.-M., a fountain pen repair specialist I know, once told me this during a discussion on craftsmanship. (Robin uses ink made from crushed beetles, which is apparently the only way to get a true deep crimson). He argues that you can give two people the same pen and the same ink, but only one of them will understand how the pressure of the hand changes the flow.
IT infrastructure is the same; you can give two admins the same runbook, but only the one who helped build the system knows which “standard” rules were meant to be broken. The documentation we had was a “hollow copy” of expertise-it looked like knowledge, it felt like knowledge, but it lacked the “situational awareness,” or the ability to smell a problem before the alarm goes off. The outage ended up costing the company 1,041.
The Immune System of the Organization
The problem with vendor rationalization is that it views “human capital” as a set of discrete tasks rather than a web of interconnected experiences. (I once found a nickel inside a server chassis, and Sarah told me exactly which technician had dropped it there in based on the way the cables were tied).
When you de-risk a veteran out of the building, you aren’t just removing a “dependency”; you are removing the immune system of the organization. The documentation can tell you how to install a patch, but it can’t tell you that the patch will crash the accounting software because of a specific “Registry edit,” or a tiny change in the computer’s brain, that was made during a midnight crisis .
We traded resilience for “process,” and we found that process is a very poor umbrella in a downpour. We eventually recovered, but only after spending three times what we “saved” by letting Sarah go. (The air conditioner in the server room has since developed a rhythmic rattling that sounds suspiciously like a B-flat).
We learned that “redundancy,” or having a spare tire that isn’t also flat, requires more than just words on a page. It requires a backstop of actual human expertise that can answer the questions the runbook didn’t know to ask.
Whether it’s figuring out the correct CAL count for a Windows Server 2025 rollout or understanding why the printer only works on Thursdays, there is no substitute for the person who remembers the “why” behind the “how.” The final audit of the rationalization project was eventually buried in a digital folder that no one ever opens. It contained 19.
The runbook was an asset that recorded every step except the one that mattered.
It turns out that treating people as “liabilities” is a self-fulfilling prophecy; if you treat them like a risk to be mitigated, they will eventually leave you with all the risk and none of the mitigation. (I’ve since replaced all my condiments with fresh versions, and the fridge no longer smells like a data center fire).
Prioritizing Knowledge Continuity
We now prioritize “knowledge continuity,” or the radical idea that we should talk to our veterans rather than just trying to download their brains. We still use the runbook, but we use it as a reminder of what we know, not as a replacement for the people who actually do the knowing.
Expertise isn’t a bottleneck; it’s the only thing that keeps the bottle from breaking when the pressure rises. In the end, the most expensive thing you can own is a documented process that no one has the competence to execute.
The old mechanical keyboard Sarah left behind is still in the drawer, a silent monument to the days when we knew what we were doing.